Changeset ff7ac6 for src/Fragmentation/Summation/Summation_impl.hpp

Timestamp:

Sep 14, 2016, 6:43:45 PM (9 years ago)

Author:

Frederik Heber <heber@…>

Branches:

Action_Thermostats, Add_AtomRandomPerturbation, Add_FitFragmentPartialChargesAction, Add_RotateAroundBondAction, Add_SelectAtomByNameAction, Adding_Graph_to_ChangeBondActions, Adding_MD_integration_tests, Adding_StructOpt_integration_tests, Automaking_mpqc_open, AutomationFragmentation_failures, Candidate_v1.5.4, Candidate_v1.6.0, Candidate_v1.6.1, ChangeBugEmailaddress, ChangingTestPorts, ChemicalSpaceEvaluator, Combining_Subpackages, Debian_Package_split, Debian_package_split_molecuildergui_only, Disabling_MemDebug, Docu_Python_wait, EmpiricalPotential_contain_HomologyGraph, EmpiricalPotential_contain_HomologyGraph_documentation, Enable_parallel_make_install, Enhance_userguide, Enhanced_StructuralOptimization, Enhanced_StructuralOptimization_continued, Example_ManyWaysToTranslateAtom, Exclude_Hydrogens_annealWithBondGraph, FitPartialCharges_GlobalError, Fix_ChargeSampling_PBC, Fix_ChronosMutex, Fix_FitPartialCharges, Fix_FitPotential_needs_atomicnumbers, Fix_ForceAnnealing, Fix_IndependentFragmentGrids, Fix_ParseParticles, Fix_ParseParticles_split_forward_backward_Actions, Fix_StatusMsg, Fix_StepWorldTime_single_argument, Fix_Verbose_Codepatterns, ForceAnnealing_goodresults, ForceAnnealing_oldresults, ForceAnnealing_tocheck, ForceAnnealing_with_BondGraph, ForceAnnealing_with_BondGraph_continued, ForceAnnealing_with_BondGraph_continued_betteresults, ForceAnnealing_with_BondGraph_contraction-expansion, GeometryObjects, Gui_displays_atomic_force_velocity, IndependentFragmentGrids_IndividualZeroInstances, IndependentFragmentGrids_IntegrationTest, IndependentFragmentGrids_Sole_NN_Calculation, JobMarket_RobustOnKillsSegFaults, JobMarket_StableWorkerPool, JobMarket_unresolvable_hostname_fix, ODR_violation_mpqc_open, PartialCharges_OrthogonalSummation, PythonUI_with_named_parameters, QtGui_reactivate_TimeChanged_changes, Recreated_GuiChecks, RotateToPrincipalAxisSystem_UndoRedo, StoppableMakroAction, Subpackage_CodePatterns, Subpackage_JobMarket, Subpackage_LinearAlgebra, Subpackage_levmar, Subpackage_mpqc_open, Subpackage_vmg, ThirdParty_MPQC_rebuilt_buildsystem, TrajectoryDependenant_MaxOrder, TremoloParser_IncreasedPrecision, TremoloParser_MultipleTimesteps, Ubuntu_1604_changes, stable

Children:

3d444a

Parents:

f856ec

git-author:

Frederik Heber <heber@…> (06/12/16 06:57:47)

git-committer:

Frederik Heber <heber@…> (09/14/16 18:43:45)

Message:

BREAKS: Summation and OrthogonalSummation both now allow to set a specific "ZeroInstance".

• this replaces then the default ZeroInstance<T> used and may be used to enforce specific parameters in the summation, e.g. a smaller grid when summing SamplingGrid.
• Note that this breaks all long-range results for the moment. The reason is that so far the ZeroInstance was effectively the first result/contribution upon which all others were summed. This worked because each fragment used the same grid size, namely the one of the global grid. With the independent fragment grids this will no longer be possible, which is why we introduce the possibility to set a specific ZeroInstance, i.e. providing the global grid to sum onto externally.
• to be absolutely precise: The reasons why the tests break is that the little change in ::Sum(): sum =.. -> sum +=..
• TESTS: Setting regression tests AnalyseFragmentResults to failed.

File:

• src/Fragmentation/Summation/Summation_impl.hpp (modified) (3 diffs)

src/Fragmentation/Summation/Summation_impl.hpp

@@ -15 +15 @@
 #endif
 
+#include "CodePatterns/Assert.hpp"
 #include "CodePatterns/IteratorAdaptors.hpp"
 #include "CodePatterns/Log.hpp"
+
+#include "Fragmentation/Summation/Summation.hpp"
 
 #include "Fragmentation/Summation/printKeyNames.hpp"
@@ -29 +32 @@
     InputValues_t& values,
     SubsetMap::ptr _subsetmap) :
-  subsetmap(_subsetmap)
+  subsetmap(_subsetmap),
+  zeroinstance(ZeroInstance<T>())
 {
   ASSERT( indices.size() == values.size(),
@@ -69 +73 @@
 {
   typename SetValueMap<T>::const_iterator iter = setvalues.begin();
-  T sum = ZeroInstance<T>();
+  T sum(zeroinstance);
   if (iter->first->size() <= level)
-    sum = (iter->second)->getValue();
+    sum += (iter->second)->getValue();
   LOG(3, "DEBUG: Value from subset "+toString(*(iter->second->getIndexSet()))
       +" is "+toString(sum)+".");